210 lines
18 KiB
Markdown
210 lines
18 KiB
Markdown
## [2026-04-15] ingest | DevOps Culture and Transformation
|
||
|
||
## [2026-04-15] ingest | 2025年11个神级AI开源平替，GitHub杀疯了
|
||
|
||
## [2026-04-15] ingest | 14个免费的AI图生视频工具，用AI让图片动起来
|
||
|
||
Added source. Key claims: DeepSeek R1是开源界首个深度推理破壁者；Qwen 3是开源界六边形战士；Flux是人体解剖学最正确的开源生图模型；HunyuanVideo是开源界参数量最大的视频生成模型；Manus定义AI Agent元年；OpenManus有5万Star；Cline是Cursor最佳开源平替；n8n有16万Star；Perplexica是完全本地化的开源AI搜索。
|
||
|
||
Created/updated: 12 entity pages (DeepSeek, Qwen, Flux, Stable Diffusion, HunyuanVideo, Manus, OpenManus, Cline, n8n, Dify, Perplexica, NotebookLM), 8 concept pages (大语言模型, AI生图, AI生视频, 通用智能体, AI编程, 智能体工作流, AI搜索, AI知识库), 1 source page.
|
||
|
||
|
||
## [2026-04-15] ingest | 3.2万人收藏的Claude Skills
|
||
- Key claims: Skills = 说明书 + SOP；从提示词工程迈向流程工程；Anthropic 官方 Skills 仓库含 3.2 万收藏；skillsmp.com/aitmpl.com/claudemarketplaces.com 三大聚合站
|
||
- Created: 1 entity (Anthropic), 2 concepts (AI技能封装, 流程工程)
|
||
|
||
## [2026-04-15] ingest | 7 Ways NotebookLM
|
||
- [7-ways-NotebookLM](sources/7-ways-NotebookLM.md)
|
||
- Key claims: Source-Grounding 机制确保回答可溯源；Audio Overviews 支持被动学习；NotebookLM 可作项目管理系统与法律文档审查工具
|
||
- Created: 2 concepts (Source-Grounding, 被动学习)
|
||
|
||
## [2026-04-15] ingest | Designing for Agentic AI
|
||
- [Designing-for-Agentic-AI](sources/Designing-for-Agentic-AI.md)
|
||
- Key claims: Agentic AI = 行动导向而非内容生成；5条设计原则：透明度、控制感、个性化、对话、预判；用户通过观察 AI 决策过程参与交互
|
||
- Created: 1 concept (Agentic AI), 1 entity (Yuri Pessa)
|
||
|
||
## [2026-04-15] ingest | LLMs、RAG、AI Agent 三个到底什么区别
|
||
- [LLMs-RAG-AI-Agent-三个到底什么区别](sources/LLMs-RAG-AI-Agent-三个到底什么区别.md)
|
||
- Key claims: LLM = 天才大脑（推理）；RAG = 随身图书馆助理（信息）；AI Agent = 行动循环（执行）；三者协同构成生产级 AI 系统
|
||
- Created: concepts (LLM, RAG, AI Agent, AI Agent 循环)
|
||
|
||
## [2026-04-15] ingest | 大模型相关术语和框架总结
|
||
- [大模型相关术语和框架总结](sources/大模型相关术语和框架总结.md)
|
||
- Key claims: LLM ≥1B 参数门槛；MCP 协议是 LLM 与外部工具的标准化接口；大模型仅输出步骤不执行；RAG 将考试正确率从 60% 提升至 90%；vLLM 通过 PagedAttention 和连续批处理优化 GPU 利用率
|
||
- Created: 8 concepts (MCP, vLLM, Token, Self-Healing Systems, RCA, Multi-Cloud Governance, 数据蒸馏), 0 entities (已有 DeepSeek/Manus/GPT-2/GPT-3)
|
||
|
||
## [2026-04-15] ingest | GitHub 上 5000 人收藏的 Vibe Coding 神级指南
|
||
- [GitHub-Vibe-Coding-神级指南](sources/GitHub-Vibe-Coding-神级指南.md)
|
||
- Key claims: Vibe Coding = 规划驱动 + 上下文固定 + AI 结对执行；开发者做导演，AI 工具承担体力活；推荐 Cursor + claude-opus-4.5-xhigh
|
||
- Created: 5 entities (Cursor, Windsurf, Trae, Karpathy, vibe-coding-cn), 1 concept (Vibe Coding)
|
||
|
||
## [2026-04-15] ingest | How Agentic AI can help for Cloud DevOps
|
||
- [How-Agentic-AI-for-Cloud-DevOps](sources/How-Agentic-AI-for-Cloud-DevOps.md)
|
||
- Key claims: Agentic AI 赋能 7 大 DevOps 场景：自主检测修复、智能 IaC、成本优化、安全合规、日志分析、多租户管理、AI 增强决策
|
||
- Created: 4 entities (AWS, GCP, Azure, Terraform, CloudWatch), 0 concepts (已有 Agentic AI)
|
||
|
||
## [2026-04-15] ingest | Multi-Agent System Reliability
|
||
- [Multi-Agent-System-Reliability](sources/Multi-Agent-System-Reliability.md)
|
||
- Key claims: 4种架构模式提升多 Agent 可靠性：Hierarchy/Consensus/Adversarial Debate/Knock-out；LLM 不可靠必须被约束验证淘汰；停止拟人化 LLM
|
||
- Created: 4 concepts (Multi-Agent Hierarchy/Consensus/Adversarial Debate/Knock-out), 1 entity (Alex Ewerlöf)
|
||
|
||
## [2026-04-15] ingest | 如何写出完美的Prompt
|
||
- [如何写出完美的Prompt（提示词）？](sources/如何写出完美的Prompt（提示词）？.md)
|
||
- Key claims: Prompt = 人与AI协作协议；Prompt能力本质是结构化思维+精准表达；建立测试-反馈-优化闭环
|
||
- Created: 3 concepts (结构化思维, 精准表达, Prompt工程)
|
||
|
||
## [2026-04-15] ingest | RAG从入门到精通
|
||
- [RAG从入门到精通系列1：基础RAG](sources/RAG从入门到精通系列1：基础RAG.md)
|
||
- Key claims: RAG = Indexing-Retrieval-Generation 三阶段；Embedding Model Context Window 512~8192 token 需文档切分；LangChain 简化 RAG 管道构建
|
||
- Created: 3 concepts (RAG, Embedding, 向量数据库)
|
||
|
||
## [2026-04-15] ingest | YouTube RSS Feed
|
||
- [How to Get the RSS Feed For Any YouTube Channel](sources/How-to-Get-the-RSS-Feed-For-Any-YouTube-Channel.md)
|
||
- Key claims: 通过 View Page Source 搜索 channel_id= 可获取 RSS Feed URL；无需第三方服务
|
||
- Created: 1 concept (RSS Feed)
|
||
|
||
## [2026-04-15] ingest | Nano Banana 提示词框架
|
||
- Key claims: 9层结构化字段（Shot/Subject/Environment/Lighting/Camera/ColorGrade/Style/Quality/Negatives）；negatives 是质量控制关键；camera 字段提供电影级构图控制
|
||
- Created: 1 concept (Nano Banana), 1 entity (Google)
|
||
|
||
## [2026-04-15] ingest | Claude Code 调用方法总结
|
||
- [Claude-Code调用方法总结](sources/Claude-Code调用方法总结.md)
|
||
- Key claims: Print Mode = stdin 管道非交互模式（推荐）；bypassPermissions 跳过所有确认；--add-dir 自动激活 SKILL；delegate_task 无法建立外部 CLI 通道
|
||
- Created: 2 concepts (Print Mode, Skill加载), 2 entities (Claude Code, Hermes)
|
||
|
||
## [2026-04-15] ingest | 万字讲透OpenClaw Workspace
|
||
- [万字讲透OpenClaw-Workspace深度解析](sources/万字讲透OpenClaw-Workspace深度解析.md)
|
||
- Key claims: AGENTS.md 300-500字最佳；SOUL vs IDENTITY 分工明确；TOOLS.md 核心是"什么时候不用"；memory/ 是真正长期记忆；7个核心文件配合实现可预期性
|
||
- Created: 7 concepts (Workspace, AGENTS.md, SOUL.md, USER.md, TOOLS.md, IDENTITY.md, BOOTSTRAP.md), 2 entities (OpenClaw, DracoVibeCoding)
|
||
|
||
## [2026-04-15] ingest | 使用Claude自动生成N8N工作流
|
||
- [使用Claude自动生成N8N工作流的实操教程](sources/使用Claude自动生成N8N工作流的实操教程.md)
|
||
- Key claims: n8n-mcp 提供 543 节点结构化访问；Claude 自动生成工作流完成度 80%-90%；Opensea 模型 + extended thinking 提升质量
|
||
- Created: 4 entities (n8n-mcp, czlonkowski, Claude, n8n), 1 concept (AI工作流自动生成)
|
||
|
||
## [2026-04-15] ingest | MCP在Cursor中的集成与应用
|
||
- [MCP在Cursor中的集成与应用详解](sources/MCP在Cursor中的集成与应用详解.md)
|
||
- Key claims: MCP 3种核心接口（GET/POST/Promise）；Cursor 支持 SSE 和 Command 两种接入方式；Agent 模式自动执行工具链；Sequential Thinking 分步推理
|
||
- Created: 4 entities (Composer, Sequential Thinking, Cursor, MCP), 3 concepts (Agent模式, MCP工具链, MCP)
|
||
|
||
## [2026-04-15] ingest | Google 5个Agent Skill设计模式
|
||
- [Google-5个Agent-Skill设计模式-2026-03-19](sources/Google-5个Agent-Skill设计模式-2026-03-19.md)
|
||
- Key claims: 5种设计模式（Tool Wrapper/Generator/Reviewer/Inversion/Pipeline）；可组合使用；Anthropic：最好的 Skill 是工具箱而非提示词
|
||
- Created: 1 entity (Google), 10 concepts (Agent Skill设计模式, Tool Wrapper, Generator, Reviewer, Inversion, Pipeline, 渐进式披露, etc.)
|
||
|
||
## [2026-04-15] ingest | A Formalization of Recursive Self-Optimizing Generative Systems
|
||
- [A-Formalization-of-Recursive-Self-Optimizing-Generative-Systems](sources/A-Formalization-of-Recursive-Self-Optimizing-Generative-Systems.md)
|
||
- Key claims: 自优化系统核心目标是生成器空间收敛而非单次最优；稳定生成能力 = Φ 的固定点；Y Combinator 实现自引用不动点；Bootstrapping = α-提示词+Ω-提示词递归超越
|
||
- Created: 3 concepts (自递归优化生成系统, 固定点, 自举), 1 entity (tukuai)
|
||
|
||
## [2026-04-15] ingest | 不会Gemini的产品经理真的要淘汰了
|
||
- [不会Gemini的产品经理真的要被淘汰了](sources/不会Gemini的产品经理真的要被淘汰了.md)
|
||
- Key claims: Gemini是知识渊博但不带脑子的苦工；FeatureList共创→Mermaid→PRD→HTML为AI产品经理核心工作流；精准表达是调教基础；超级个体=某领域八九十分+AI放大
|
||
- Created: 4 concepts (FeatureList, PRD自动生成, AI产品经理, 超级个体), 2 entities (Kira2red, Gemini)
|
||
|
||
## [2026-04-15] ingest | baoyu-skills 宝玉 Claude Code 技能集
|
||
- [baoyu-skills-claude-code技能集](sources/baoyu-skills-claude-code技能集.md)
|
||
- Key claims: ClawHub按单个skill安装；内容技能覆盖xhs/信息图/幻灯片/漫画；baoyu-imagine支持9家服务商自动选择；工具技能覆盖翻译/字幕/URL抓取等
|
||
- Created: 6 concepts (baoyu-imagine, baoyu-infographic, 内容技能, AI生成技能, 工具技能, ClawHub), 2 entities (宝玉, JimLiu)
|
||
|
||
## [2026-04-15 PM] ingest batch | 4 docs (n8n Docker / Cloud Operating Model / MinIO+Zipline / Trae)
|
||
- [n8n-Docker安装与更新](sources/n8n-Docker安装与更新.md)
|
||
Key claims: n8n 官方镜像需 Dockerfile 扩展安装 curl/wget；ALL_PROXY=socks5://172.21.0.1:10808 使容器内流量走宿主机代理；更新流程：pull && down && up -d
|
||
Created: 1 concept (Docker容器网络)
|
||
|
||
- [Cloud-Operating-Model-关键策略与最佳实践](sources/Cloud-Operating-Model-关键策略与最佳实践.md)
|
||
Key claims: COM 四大支柱（治理/自动化/安全/FinOps）；Zero Trust=零隐式信任持续验证；FinOps 可降低 35-40% 云成本；六步设计法
|
||
Created: 4 concepts (Cloud Operating Model, FinOps, Zero Trust, 云治理)
|
||
|
||
- [MinIO-Zipline-自托管图床应用安装教程](sources/MinIO-Zipline-自托管图床应用安装教程.md)
|
||
Key claims: MinIO+Zipline+PostgreSQL 三件套 Synology Docker 部署；pg_dump 热备份+Hyper Backup 增量归档防数据不一致；mc anonymous set public 设置公开 Bucket
|
||
Created: 2 entities (MinIO, Zipline), 4 concepts (S3协议, Docker Compose, PostgreSQL备份, Synology Hyper Backup)
|
||
|
||
- [Trae远程开发部署指南](sources/Trae远程开发部署指南.md)
|
||
Key claims: Remote SSH 连接 Ubuntu 服务器；Attach 容器模式=隔离环境；Bind Mount=热更新；SSH Agent 转发解决 Git 凭证问题
|
||
Created: 3 concepts (Remote SSH, Docker Attach模式, Bind Mount)
|
||
|
||
## [2026-04-15] ingest batch | 5 docs
|
||
- [Claude Skills研究范式](sources/Claude-Skills研究范式.md)
|
||
- Key claims: Skills = 说明书 + SOP；从提示词工程迈向流程工程；Anthropic 官方 Skills 仓库含 3.2 万收藏；skillsmp.com/aitmpl.com/claudemarketplaces.com 三大聚合站
|
||
- Created: 1 entity (Anthropic), 2 concepts (AI技能封装, 流程工程)
|
||
|
||
- [递归自优化生成系统的形式化框架](sources/递归自优化生成系统的形式化框架.md)
|
||
- Key claims: 自优化目标 = 生成器空间 {G_n} 的收敛行为；稳定生成能力 = Φ 的固定点 G*；Y Combinator 表达 G* = Y STEP；自举实现自我超越
|
||
- Created: 3 concepts (自递归优化生成系统, 固定点, Y Combinator), 1 entity (tuuai)
|
||
|
||
- [AI解决方案专家培训课程](sources/AI解决方案专家培训课程.md)
|
||
- Key claims: Coze 平台（coze.cn/coze.com）覆盖 8 大行业 Agent Demo；支持单 Bot 对话、Workflow 工作流、表格问答等多种类型
|
||
- Created: 1 entity (Coze)
|
||
|
||
- [NotebookLM的7种用法](sources/NotebookLM的7种用法.md)
|
||
- Key claims: Source-Grounding 机制确保回答可溯源；Audio Overviews 支持被动学习；NotebookLM 可作项目管理系统、法律文档审查工具、软件更新对比工具
|
||
- Created: 1 entity (NotebookLM), 2 concepts (Source Grounding, 被动学习)
|
||
|
||
- [Nano Banana结构化提示词框架](sources/Nano-Banana结构化提示词框架.md)
|
||
- Key claims: 9层结构化字段（Shot/Subject/Environment/Lighting/Camera/ColorGrade/Style/Quality/Negatives）；negatives 是质量控制关键；camera 字段提供电影级构图控制
|
||
- Created: 1 concept (Nano Banana, 负向提示词), 1 entity (Google)
|
||
|
||
## [2026-04-15] ingest batch | 3 docs (Home Office + Cloud DevOps)
|
||
|
||
- [Cloud-DevOp-Maturity-Guideline](sources/Cloud-DevOp-Maturity-Guideline.md)
|
||
- Key claims: DevOps 成熟度评估框架；DORA 四指标（部署频率、变更前置时间、变更失败率、平均恢复时间）；四大支柱（自动化、协作与文化、监控与可观测性、DevSecOps）
|
||
- Created: 1 concept (DORA指标)
|
||
|
||
- [Cloud-Maturity-Model](sources/Cloud-Maturity-Model.md)
|
||
- Key claims: 云成熟度 5 级模型（Legacy→Initial→Repeatable→Systematic→Measured→Optimized）；三要素（People/Processes/Technology）；Forrester 预测2025年市场达15亿美元
|
||
- Created: concepts (云成熟度模型, 云迁移, 云治理, CAPEX, OPEX, TCO)
|
||
|
||
- [Linux-运维必会的150个命令](sources/Linux-运维必会的150个命令.md)
|
||
- Key claims: 150 个 Linux 核心命令速查；一切皆文件理念；按功能分类（线上查询、文件操作、文本处理、信息显示、压缩解压）
|
||
- Created: concepts (Shell, 管道, 正则表达式)
|
||
|
||
## [2026-04-15] ingest batch | 3 docs (监控方案 + NAS媒体平台 + 电商爬虫)
|
||
|
||
- [家庭监控方案 Prometheus + Grafana + Node Exporter + cAdvisor + Blackbox](sources/家庭监控方案 Prometheus + Grafana + Node Exporter + cAdvisor + Blackbox.md)
|
||
- Key claims: Prometheus pull模式采集主机/容器/服务指标；Grafana Dashboard ID 1860/14282/7587；docker-compose 一键部署完整监控栈；Docker Socket 挂载存在安全风险；TLS 证书到期可通过 probe_ssl_earliest_cert_expiry 监控
|
||
- Created: entities (Prometheus, Grafana, Loki, Uptime Kuma), concepts (可观测性, 合成监测)
|
||
|
||
- [Synology NAS + Xiaoya Alist + CloudDrive2 + Plex to Build Media Platform](sources/Synology NAS + Xiaoya Alist + CloudDrive2 + Plex to Build Media Platform.md)
|
||
- Key claims: 离线镜像导入法（docker save/load）解决 Container Manager 无法访问 Docker Hub；Xiaoya Alist 需 myopentoken/mytoken/temp_transfer_folder_id 三个配置文件；CloudDrive2 需 DSM 7+ privilege 修复；Plex 通过 TMDB/TVDB 自动刮削
|
||
- Created: entities (Plex, Synology NAS, 阿里云盘), concepts (媒体刮削)
|
||
|
||
- [可自动化可扩展AI增强的电商数据采集与处理系统](sources/可自动化可扩展AI增强的电商数据采集与处理系统.md)
|
||
- Key claims: Scrapy + scrapy-playwright 处理静态+动态页面；n8n Workflow 自动化管线（Cron→爬虫→AI→存储→通知）；Ollama 本地模型通过 HTTP API 调用；防封策略（UA轮换/代理池/延迟访问）
|
||
- Created: entities (Scrapy, Playwright, Ollama), concepts (电商数据采集, AI数据处理)
|
||
|
||
## [2026-04-15 PM] ingest batch | Multi-Agent Team / DevOps Maturity / 一语点醒梦中人 / NodeWarden
|
||
- [Agent-usecases-multi-Agent-Team](sources/Agent-usecases-multi-Agent-Team.md)
|
||
Key claims: 单一 Agent 无法多领域高效工作；共享记忆+私有上下文是多 Agent 协作核心；定时主动任务是价值飞轮；从 2 Agent 开始按瓶颈扩展
|
||
Created: 2 entities (Trebuh, Cloudflare), 3 concepts (DevOps成熟度模型, 共享内存模式, 空性智慧, 绝处逢生)
|
||
- [DevOps-Maturity-Model-From-Traditional-IT-to-Advanced-DevOps](sources/DevOps-Maturity-Model-From-Traditional-IT-to-Advanced-DevOps.md)
|
||
Key claims: DevOps 成熟度 5 阶段（Ad-Hoc→Pockets→Defined→Optimized→Mature）；4大焦点：文化/自动化/结构/协作；DORA 四指标量化效能
|
||
- [AI-一语点醒梦中人](sources/AI-一语点醒梦中人.md)
|
||
Key claims: 王维"行到水穷处"体现佛学超脱；曾国藩"忘机消众机"融合道家无为；"知其不可奈何而安之若命"区分可控与不可控
|
||
- [Home-Office-NodeWarden-把-Bitwarden-搬上-Cloudflare-Workers彻底告别服务器](sources/Home-Office-NodeWarden-把-Bitwarden-搬上-Cloudflare-Workers彻底告别服务器.md)
|
||
Key claims: NodeWarden 将 Bitwarden 托管在 Cloudflare Workers；支持 TOTP（官方需会员）；零服务器费用；Cloudflare D1+R2 提供存储
|
||
|
||
## [2026-04-15] ingest | Last30Days 使用指南
|
||
- [Last30Days-使用指南](sources/Last30Days-使用指南.md)
|
||
- Key claims: 8 大数据源热点聚合（Reddit/X > YouTube > Polymarket > TikTok > Instagram > Web）；社交信号权重基于真实互动而非曝光量；对比模式一次返回 A/B 并排研究；深度研究需 2-8 分钟
|
||
- Created: 1 entity (ScrapeCreators), 2 concepts (多平台热点聚合, 社交信号权重)
|
||
|
||
## [2026-04-15] ingest | GOG CLI 安装配置指南
|
||
- [GOG-CLI-安装配置指南](sources/GOG-CLI-安装配置指南.md)
|
||
- Key claims: macOS Google Workspace CLI；OAuth 双层验证（凭证 + API Enablement）；403 accessNotConfigured 根因是 API 未启用；添加测试用户绕过验证限制
|
||
- Created: 1 entity (gog CLI), 1 concept (Google Workspace CLI)
|
||
|
||
## [2026-04-15] ingest | Cursor 2.0 初学者使用指南
|
||
- [Cursor-2-0初学者使用指南](sources/Cursor-2-0初学者使用指南.md)
|
||
- Key claims: Composer 模型比同类快 4 倍；多代理并行（Plan/Agent/Ask 三模式）；Diff 审查强制先测试再确认；AI 生成代码即写入文件
|
||
- Created: 1 entity (Cursor), 4 concepts (AI代码编辑器, Composer模型, 多代理并行, Diff审查, 项目规则)
|
||
|
||
## [2026-04-15 PM] ingest | Claude Skills 完整版 + Designing Agentic AI + NotebookLM 7场景 + 养虾日记1 + AI时代赚钱
|
||
- [3.2万人收藏的Claude Skills（完整版）](sources/3.2万人收藏的Claude-Skills才是AI这条路最值得研究的一套范式.md)
|
||
- [Designing for Agentic AI](sources/Designing-for-Agentic-AI.md)
|
||
- [7 Ways I Use NotebookLM](sources/7-ways-NotebookLM.md)
|
||
- [养虾日记1：用 OpenClaw 管了 28 万张照片](sources/养虾日记1-OpenClaw照片整理实战.md)
|
||
- [不谈技术：普通人该怎么在AI时代赚钱](sources/普通人如何在AI时代赚钱.md)
|
||
- Created entities: ComposioHQ, VoltAgent, BehiSecc, Yuri Pessa
|
||
- Created concepts: 流程工程, 透明度, 控制权, 个性化, 对话式设计, 预判式设计, 精确去重, 小文件清理, 安全删除, 分批执行, AI Agent 思维方式, 品味, 端到端, 死亡过滤器, 引文追溯, 被动学习
|